Inproceedings,
Recoverable Team Formation: Building Teams Resilient to Change
Proceedings of the 17th International Conference on Autonomous Agents and MultiAgent Systems, page 1362–1370. Richland, SC, International Foundation for Autonomous Agents and Multiagent Systems, (2018)
Abstract
Team formation consists in finding the least expensive team of agents such that a certain set of skills is covered. In this paper, we formally introduce recoverable team formation (RTF), a generalization of the above problem, by taking into account the dynamic nature of the environment, e.g. after a team has been formed, agents may unexpectedly become unavailable due to failure or illness. We analyze the computational complexity of RTF, provide both complete and heuristic algorithms, and empirically evaluate their performance. Furthermore, we demonstrate that RTF generalizes robust team formation, where the task is to build a team capable of covering all required skills even after any k agents are removed. Despite the high complexity of forming a recoverable team, we argue that recoverability is a crucial feature, and experimentally show that it is more appropriate for some applications than robustness.
